Professional Documents
Culture Documents
Research Paper
On the application of the beam model for linear dynamic analysis of pile
and suction caisson foundations for offshore wind turbines
Guillermo M. Álamo *, Jacob D.R. Bordón , Juan J. Aznárez
Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería (SIANI), Universidad de Las Palmas de Gran Canaria, Edificio Central del Parque
Científico y Tecnológico, Campus Universitario de Tafira, 35017 Las Palmas de Gran Canaria, Spain
A R T I C L E I N F O A B S T R A C T
Keywords: Piles and suction caissons are the most common foundation solutions for fixed Offshore Wind Turbines at in
Offshore wind turbine termediate water depths. They are generally used as a single element, presenting large diameters and short aspect
Suction caisson foundation ratios. These specific dimensions drastically differ from the ones of classical applications (offshore platforms,
Monopile foundation
bridges, tall buildings etc.). Thus, in this paper the validity of their modelling as beam elements for the particular
Seismic response
problem of OWT is revised. The results of a soil-beam model, based on the integral Reciprocity Theorem in
Boundary elements
Finite elements Elastodynamics and specific Green’s functions for the layered half-space for the soil behaviour coupled with
Timoshenko’s beam Finite Elements, are benchmarked against the ones of a soil-shell model, based on Boundary
Elements for the soil coupled with shell Finite Elements. The comparative study is conducted in terms of
foundation characterization variables (impedance functions and kinematic interaction factors). Their influence
on the OWT seismic response is also studied through a substructuring procedure. From the results, some ex
pressions for determining the applicability range of the beam simplification are proposed as functions of the
relative foundation-soil stiffness ratio. It is observed that this applicability range goes beyond that the one
commonly considered. .
1. Introduction (pile shaft) for monopiles and t/D ∼ 0.001 (caisson skirt) for suction
caissons.
Pile and suction caisson foundations (also known as buckets, suction The complete design of an OWT is a complex task which should fulfil
piles or suction anchors depending on the context) are being used as many requirements (DNV, 2014). Although rough designs can be ob
foundations of fixed Offshore Wind Turbines (OWT). Such solutions are tained from simplified procedures at early stages, see e.g. (Arany et al.,
being considered at sites with shallow (10 to 30 meters) and interme 2017), final designs should be defined after using an integrated opti
diate water depths (30 to 60 meters) with single (monopile or mono mization procedure which guarantees a safe and economical solution,
bucket) or multiple foundation arrangements depending on both water see e.g. (Ashuri et al., 2014). In this sense, it is of fundamental impor
depth and soil properties. tance an appropriate modelling which should also be computationally
As wind turbines becomes larger, the required piles and suction efficient due to the many evaluations needed during the design process.
caissons becomes bigger in diameter: up to 8 meters for monopiles (XXL In the case of foundation modelling, recent works (Bhattacharya et al.,
monopiles), up to 30 meters for caissons (Cotter, 2009), and even bigger 2013; Bordón et al., 2019; Page et al., 2019), have highlighted its
for shallow composite caissons with internal skirts (see e.g. Jia et al., importance particularly for the prediction of first and second natural
2018). Such diameters are much bigger than diameters used in other frequencies, which are relevant for performing the crucial fatigue ana
more classical projects (gas/oil platforms, bridges, buildings, etcetera), lyses. In the case of seismic analyses, the frequency range usually goes
and therefore the use of models traditionally considered should be taken far beyond the second natural frequency. The use of convenient
with care. The length L to diameter D ratios ranges are 3 < L/D < 10 for simplifying hypotheses allows adopting simpler and cheaper models
monopiles and 1 < L/D < 6 for suction caissons (Houlsby and Byrne, (ideally via closed-form formulae or calibrated fast models) at initial
2005a; Houlsby and Byrne, 2005b). The buried part of both piles and steps, whereas more rigorous, complex and detailed models should be
caissons are constituted by steel shells of thickness around t/D ∼ 0.01 considered at the final stages.
* Corresponding author.
E-mail addresses: guillermo.alamo@ulpgc.es (G.M. Álamo), jacobdavid.rodriguezbordon@ulpgc.es (J.D.R. Bordón), juanjose.aznarez@ulpgc.es (J.J. Aznárez).
https://doi.org/10.1016/j.compgeo.2021.104107
Received 7 October 2020; Received in revised form 25 January 2021; Accepted 25 February 2021
Available online 21 March 2021
0266-352X/© 2021 Elsevier Ltd. All rights reserved.
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
The present paper is concerned with the modelling of big diameter interaction (impedances and kinematic interaction factors) for length
and short piles and suction caissons for seismic analyses of OWT sup to diameter ratios as low as 1, i.e. suction caissons.
ported by a single foundation element (monopile or monobucket). The The paper is organized as follows. First, Section 2 presents the two
focus is put on assessing the adequacy of different simplifying hypothesis different foundation models together with the substructuring model
which allow the use of more elementary models from which results are used to compute the OWT system response. Then, the comparison be
more easily obtained. To this end, a previously developed soil-beam tween the beam and shell models is conducted along the different parts
coupled model (Álamo et al., 2016) is compared against a rigorous of Section 3. The comparison starts in terms of the foundation response:
soil-shell model (Bordón et al., 2017). The former is a model which re static (Section 3.1) and and dynamic (Section 3.2) stiffness, and kine
duces the soil-structure interaction to a line (pile axis), and greatly re matic interaction factors (Section 3.3). Then, some expressions deter
duces the number of degrees of freedom. The pile itself is modelled with mining the upper frequencies below which the beam model accurately
Timoshenko finite elements, and the soil response is included via reproduces the foundation response are proposed in Section 3.4. The
Green’s function for an arbitrary horizontally layered half-space, which final part of the comparison, in terms of OWT variables, is done through
avoid the discretisation of the free-surface and the layer interfaces, i.e. the application example presented in Section 3.5. The paper ends listing
only pile axis nodes are present. The latter model reduces the soil- the main conclusions drawn from the study in Section 4.
structure interaction to the shell mid-surface, which leads to a
rigorous and general interaction model but it is also computationally 2. Methodology
costlier. The comparative study is performed in terms of impedances and
kinematic interaction factors. In addition to this, results in terms of OWT 2.1. Substructuring model
variables are also compared in order to evaluate to what extent the
differences in the foundation modelling are transmitted to the whole The problem at hand is summarized in Fig. 1: an Offshore Wind
tower-supporting structure-foundation system. Turbine connected to a submerged structure founded on a large diam
The use of some kind of one-dimensional reduction, i.e. beam-like eter and relatively short monopile or suction caisson subjected to
models, for foundation modelling is quite advantageous for obvious vertically-propagating S waves.
mathematical and computational reasons. The key of such models is how A two-dimensional (lateral behaviour) substructuring model is used
soil-structure interaction is considered. Pioneering works take the for modelling the OWT dynamics. It comprises a simple concentrated
rigorous Mindlin’s solution and integrate it along the pile axis or pile mass at the hub representing the Rotor-Nacelle-Assembly (RNA), a suf
shaft under certain assumptions (see e.g. Jiménez-Salas and Belzunce, ficient number of Euler–Bernoulli beam elements for taking into account
1965; Thurman and D’Appolonia, 1965; Poulos and Davis, 1968). An the conical tower and the submerged part. The foundation dynamics is
alternative approach is the well-known Winkler model, where distrib synthesized via frequency dependent springs and dashpots (impedance
uted springs and dashpots are connected along the pile axis (Novak functions) connected to the submerged part base. The foundation energy
et al., 1978; Gazetas and Dobry, 1984). Advanced approaches use some filtering when impinged by a seismic action (vertically incident shear
Green’s function under point or ring loading to introduce the soil waves) is taken into account via kinematic interaction factors. An
response, and then this is coupled to beam finite elements in different adequate representation of both impedances and kinematic interaction
ways (Kaynia and Kausel, 1982; Coda et al., 1999; Almeida and de Paiva, factors is vital for the study of OWT modal and seismic behaviour.
2004; Padrón et al., 2007). The soil-beam model used in this paper ex The impedance functions relate the forces and moments and the
tends (Padrón et al., 2007) by considering an arbitrary horizontally displacements and rotations of the foundation at the mudline level. They
layered half-space and Timoshenko finite elements. The aim of the are frequency-dependent complex functions whose real and imaginary
present work is to study to what extent this soil-beam model, initially parts represent respectively the stiffness and damping characteristics of
developed for pile foundations, is able to reproduce soil-structure the foundation. Given that the study focuses on the lateral behaviour,
2
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
Fig. 2. Foundation configuration and its modelling: (a) steel hollow pile in homogeneous half-space, (b) DBEM-FEM model (Bordón et al., 2017) (soil-shell inter
action), (c) Integral model (Álamo et al., 2016) (soil-beam interaction).
3
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
∫ ∫
1 ( i+ ) velocity, nj is the unit normal at the boundary point, and eiωt has been
t − ti−l + s*lk uk dΓ = d*lk tk dΓ (6)
2 l Γ Γ omitted for brevity.
where ulk denotes the shell displacements and tlk the distributed mid- where ̃
*
ulk is the displacement Green’s function for the layered half space
surface shell load. proposed by Pak and Guzina (Pak and Guzina, 2002) and qk are the
The seismic input is included in the formulation by following the distributed interaction forces acting over the soil. In order to numeri
classical decomposition of the total field into the superposition of the cally evaluate the line integral, classic lineal (2 nodes) elements are
incident field (produced by the impinging seismic excitation) and the considered. Also, a particular non-nodal collocation strategy over four
scattered field (produced by the foundation), see e.g. (Domínguez, points of the fictitious soil-shell interface is required in order to avoid
1993): the singularity of the Green’s function (see Álamo et al., 2016 for more
utot in sc
(9a) details).
k = uk + uk
The finite element modelling of the foundation piece is done by using
ttot in sc
(9b) two-noded beam elements. Cubic and quadratic shape functions that
k = t k + tk
satisfy the Timoshenko’s beam static equation (Friedman and Kosmatka,
Since the displacements and tractions present in the BIEs have to be 1993) are used for the lateral behaviour, while linear shape functions are
evanescent, they are substituted by the scattered field, which, under the used to model the distributed interaction forces and axial displacements.
previous assumption, is equivalent to subtracting the incident field from The equilibrium equation for a given beam element l can be written as:
the total field. The considered incident field is a simple vertically inci
dent shear wave, which has the following non-zero displacements and
̃ (l) a(l) − Q(l) q(l) = f (l)
K (12)
tractions:
Note that the terms of this equation are the beam-counterparts of the
1( ) ones presented in Eq. (7), being q(l) the vector defining the nodal values
uin = e− iks x3
+e iks x3
(10a)
1
2 of the distributed interaction forces acting over the foundation element.
Conforming meshes are considered to discretize the soil load line and
tin in
1 = Gs u1,3 n3 (10b) the foundation element. Thus, the coupling between both regions can be
easily done by, again, imposing compatibility and equilibrium condi
tin in
3 = Gs u1,3 n1 (10c) tions in terms of displacements and interaction forces, respectively:
√̅̅̅̅̅̅̅̅̅̅̅̅ uk = ulk (13a)
where ks = ω/cs is the S wavenumber, cs = G*s /ρs is the shear wave
4
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
shell (bucket skirt or pile shaft) and the soil can effectively be synthe
qk + qlk = 0 (13b)
sized via a dimensionless parameter JDoherty = (Ef t)/(Gs R) which relates
shell membrane stiffness and soil stiffness (Ef is the Young’s modulus of
where ulk denotes the beam displacements and qlk the distributed inter
the foundation and R is the radius of the foundation cross-section). It
action forces acting over the beam.
allows an approximate but useful reduction of the number of defining
Finally, and following the same strategy than in the previous soil-
parameters. For convenience’s sake, this dimensionless parameter is
shell model, the seismic excitation is introduced by superposing the
defined as J = (Gf t)/(Gs D) in the present paper, which differs from the
incident and scattered fields. Note that, owing to the reduced expression
original by the factor JDoherty /J = 4(1 +νf ) (constant since νf = 0.25 in all
of the Singular BIE, in the soil-beam model only the displacement terms
cases). Although the work of Doherty et al. (2005) is limited to static
of the incident field are necessary.
stiffnesses, it is shown in Appendix A that, in general, J remains as a
useful parameter in dynamics for frequencies ao = fD/cs < 0.3 and
3. Results and discussion t/D ≤ 0.05.
The rest of dimensionless parameters that define the studied prob
In this section, a comprehensive comparative study between results lems are: hysteretic damping ratios of foundation ξf = 2% and soil ξs =
from both models is given. It includes the necessary ingredients for 5% materials, foundation soil density ratio ρf /ρs = 3.9, and soil Pois
comparing the foundation characterization (impedances and kinematic son’s ratio νs = 0.49 (saturated soil).
interaction factors) as well as the final OWT variables of interest
(bending moments, shear forces, displacements and rotations).
The study covers the following parameters: length to diameter ratio: 3.1. Static stiffnesses
L/D = {1, …, 10}, thickness to diameter ratio: t/D =
{0.001, 0.01, 0.02}, and foundation shear modulus to soil shear Static stiffnesses are relevant for the calculation of the fundamental
modulus ratio: Gf /Gs = {1000, 4000, 16000}. frequency of OWTs since it is usually small, typically f1 ∼ 0.3 Hz (Kay
For this class of foundations in homogeneous soils this set of pa nia, 2018). For seismic analyses, however, the use of static stiffnesses is
rameters completely defines the problem. Nonetheless, Doherty et al. generally not recommended.
(2005) found that the relative stiffness between the foundation lateral Fig. 3 shows the vertical, horizontal, rocking and sway-rocking cross-
5
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
coupling stiffness components for most of the cases studied. One case per decreases as J increases, except for rocking and sway-rocking cross-
t/D has been removed since they have similar values of J (and similar coupling stiffnesses for L/D = 2. The vertical stiffness shows a gradual
graphs) to other cases. This way, the stiffness components are distrib reduction of the error as L/D increases, whereas the lateral mode stiff
uted along columns, and the relative stiffnesses between foundation and nesses show little differences in the behaviour for L/D ≥ 4. Overall, the
soil are distributed along rows. Each graph shows L/D in abscissas, and beam model achieves errors below 5% for the lateral mode stiffnesses
contains the results from both models. In addition to the aforementioned and errors below 10% for the vertical stiffness when J ≥ 10 and L/D ≥ 4.
Poisson’s ratio νs = 0.49 (saturated soil), a value of νs = 0.3 has been
also considered in this section because the soil is expected to behave 3.2. Impedance functions
more similar to the drained solid in the static regime. However, the
influence of the Poisson’s ratio in the comparison between the two The use of dynamic stiffnesses (also known as impedance functions)
models is negligible. The beam model leads to similar results to the shell allows a much more general linear SSI analysis. They can be used for the
model except for very small values of J. This phenomenon is reasonable calculation of OWT natural frequencies, as well as seismic analyses via
since the validity of the beam–soil continua is kept as long as the the substructuring procedure. In this section, the comparison between
foundation behaves as a structural member, which happens when there beam and shell models is extended to impedance functions. The
exists stiffness contrast between this element and the soil. On the other frequency-dependent differences between the two models are quantified
hand, the applicability of the beam model regarding the length to through the relative error defined in Eq. (14). The proposed expression
diameter ratio (L/D) is surprisingly good even at L/D = 1. compares the absolute value of the complex difference between the
In order to give a measure of the fidelity of the soil-beam model, the result X obtained by the shell or beam approaches with respect to the
relative error between both models is given in Fig. 4. Each graph has maximum value obtained by the reference model (i.e., soil-shell
now the parameter J in abscissas, so that all different cases of t/D and approach) in the studied frequency range. This definition is preferred
Gf /Gs are represented. Each length to diameter ratio is represented using over a frequency-by-frequency relative comparison in order to avoid
a different color: L/D = 2 (black), L/D = 4 (red), L/D = 6 (blue), L/D = peak values of the error around the frequencies in which the reference
8 (green) and L/D = 10 (orange); and each thickness to diameter ratio is result approaches to zero.
shown with a different point type: t/D = 0.001 (plus), t/D = 0.01 (cross)
and t/D = 0.02 (square). In all cases, it is roughly observed that error
Fig. 5. Comparison between dynamic stiffnesses: L/D = 6, t/D = 0.01 and Gf /Gs = 4000 (J = 40).
6
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
vious illustrative case) for all cases under study. The line and point styles
are also similar to the preceding section. The previously mentioned limit
of ao ≥ 0.25 (associated with a quarter-wavelength per foundation
diameter) is a valid indicative value for obtaining errors below 5%,
although only when J ≥ 10. In the vertical and rocking components,
o,5% are located below 0.2. This erratic behaviour is nonetheless
some alim
due to the presence of a peak in the beginning of these error curves
which lightly exceed the 5% (see supplementary data). This is further
demonstrated by observing that alim o,10% do not show this behaviour.
Fig. 5 shows the impedance curves (real and imaginary parts) as well
as the relative error for the illustrative case with L/D = 6, t/D = 0.01
and Gf /Gs = 4000, i.e. J = 40. All other cases are included as supple
mentary data. For all stiffness components, the error between the beam
and the shell models is approximately constant initially, and then it
starts to increase with the frequency. Such effect is physically justifiable
in terms of the comparison between soil wavelength and foundation
diameter. For frequencies leading to wavelengths comparable to the
foundation diameter, the reduction to a load line performed by the beam
model becomes inadequate. Thus, it is reasonable that this model starts Fig. 8. Limiting frequencies alim
o,5% (error <5%) and ao,10% (error <10%) of ki
lim
7
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
Table 1 o,10% ∼ 0.15 ÷ 0.5. There is no clearly defined trend for the limiting
alim
Coefficients for the expressions of the limiting frequencies. Error <5%. frequencies due to the presence of the previously mentioned peaks.
c0 c1 c2 Not applicable for Nonetheless, it is reasonable to use the conventional quarter-wavelength
limiting frequency (alim
o = 0.25) for the integral model regarding kine
KV 0.2584 − 0.0332 0.0054 J < 4, L/D < 4
KH 0.0322 0.0881 − 0.0111 J<4
matic interaction factors, which achieves errors up to approximately
KR 1.2758 − 0.4636 0.0455 J < 16, L/D < 4, Gf /Gs < 4000
10% for L/D > 2 and J > 10.
Iu 0.1135 − 0.0282 0.0065 L/D < 6 (use alim
o ≈ 0.084)
3.4. Expressions for the limiting frequencies
Iθ 0.0343 0.0157 0 L/D < 6 (use alim
o ≈ 0.066)
The limiting frequencies are roughly alim the last column of each table. The proposed formulas (lines) together
o,5% ∼ ao,10% ∼ 0.1 for L/D = 2,
lim
with the previous results (points) are shown in Fig. 9. In each graphical
whereas for L/D > 2 and J > 10 these are alim
o,5% ∼ 0.1 ÷ 0.2 and area, the limiting frequency is plotted against the dimensionless
Table 2
Coefficients for the expressions of the limiting frequencies. Error <10%.
c0 c1 c2 Not applicable for
KV 0.2664 0.0041 0 -
KH 0.2760 0 0 J<4
KR 0.9667 − 0.2693 0.0262 J < 10, Gf /Gs < 4000
Iu 0.5248 − 0.0462 0.0026 L/D < 6 (use alim
o ≈ 0.13 [L/D = 2] or 0.32 [L/D = 4])
Iθ 0.0414 0.0335 0 L/D < 4 (use alim
o ≈ 0.10)
Fig. 9. Comparison between the obtained limiting frequencies (points) and their proposed expressions (lines).
8
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
Fig. 10. Frequency Response Functions for representative variables of the OWT system. Comparison between the different models.
parameter J. Each column corresponds to a different foundation vari thickness ratio (t/D = 1%) as the supporting structure. An intermediate
able, while each row corresponds to a different maximum permitted embedment length ratio L/D = 6 is assumed in this example. The whole
error. Note that the points outside the applicability range of the obtained structure (tower, supporting monopile and foundation) is considered to
expressions are not included in the figure. For the kinematic interaction be made of steel with: Young’s modulus Ef = 210 GPa, density ρf = 7850
factors, the alternative values for the points outside the applicability kg/m3 and Poisson’s ratio νf = 0.25. For the superstructure a hysteretic
range are also shown in red and purple colours. damping coefficient ξf = 2.5% is considered.
The soil properties are selected in order to reproduce a saturated
3.5. Application example media through elastic equivalent properties: density ρs = 2000 kg/m3 ,
Poisson’s ratio νs = 0.49 and hysteretic damping coefficient ξs = 5%.
In order to illustrate the accuracy of the soil-beam model to repro Different foundation-soil stiffness ratios J = 10, 40 and 160 are assumed
duce the response of the OWT-foundation system, a practical example is in order to cover the range studied in the previous sections. These values
briefly presented in this section. The system is based on the reference correspond to soils whose shear wave velocities are approximately cs =
NREL-5 MW OWT model (Jonkman et al., 2009). The tower is 70 m high 200, 100 and 50 m/s, respectively.
and it has a hollow cross-section with variable diameter from 6 m at its In coherence with the obtained kinematic interaction factors, the
base to 3.87 m at the hub height, whereas a constant thickness to considered seismic excitation is a vertically incident shear wave. The
diameter ratio of 0.45% is assumed. The supporting structure is a 20 m free-field displacement is denoted as uff , while its acceleration is üff .
high, 6 m diameter and thickness ratio of 1% tubular member. No Fig. 10 shows the Frequency Response Functions (FRF) of several
transition piece between supporting structure and tower is considered in representative variables of the OWT system. The second and third rows
the analyses. The foundation presents the same diameter (D = 6 m) and present the FRF of the displacement and rotation atop the turbine tower
9
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
Fig. 11. Validity of J for characterizing the relative foundation-soil stiffness in the computation of impedance functions.
with respect to the free-field displacement. On the other hand, the fourth are not significant for the example case studied (note the logarithmic
and fifth rows illustrate the shear force and bending moment at the base scale in the forces at the base variables). Thus, the results show that the
of the supporting structure (mud-line level) with respect to the free-field real factor that limits the use of the soil-beam model is the maximum
acceleration. The results obtained by assuming a fixed-base model (no value of the dimensionless frequency ao = 0.5, i.e., foundation diameter
SSI) are compared with the ones obtained through the substructuring equal to half of the wavelength. This restriction makes the soil-beam
procedure (see Section 2.1) using both the soil-shell or soil-beam models model to be used with caution for seismic analyses of large diameter
to characterize the foundation. In order to help in interpreting the re foundations in extremely soft soils. The validity of the beam model in
sults, the first row shows the limiting frequencies (corresponding to a this scenario will strongly depend on the frequency content of the
maximum relative error between the shell and beam models less than excitation, and the evolution with frequency of the studied FRF.
10%) for all of the foundation variables involved in the problem at hand.
Each column of the figure corresponds to a different soil-foundation 4. Conclusions
relative stiffness ratio. Frequencies up to 20 Hz are considered as a
wide frequency range for the energy content of the seismic excitation. This paper presents a comparison between soil-shell and soil-beam
The results presented in Fig. 10 show a clear influence of the SSI models for the dynamic characterization of OWT foundation elements
effects on all studied variables. In general terms, introducing the foun (piles or suction caissons) and the seismic analysis of the complete
dation behaviour results in a higher response of the system for small system. The reference results are computed with the first model, that
frequencies (< 5 Hz), and a lower response for higher frequencies. The uses Boundary Elements to model the soil behaviour coupled with shell
shifting of the system natural frequencies toward lower values produced Finite Elements to represent the foundation element. On the other hand,
by the SSI effects is also seen in the obtained results. the soil-beam model is based on the integral expression of the Reci
The accuracy of using the beam model instead of the shell one can be procity Theorem together with advanced Green’s functions for the soil
also tested by comparing their results in Fig. 10. A good agreement modelling, while the structural behaviour of the foundation is handled
between both models is found for all variables. Some discrepancies are via beam Finite Elements. For the analyses, soil and foundation
produced for frequencies above certain limiting frequencies. The hori geometrical and material properties typical of this singular construction
zontal impedance term seems to be the most important one, followed by are assumed.
the limit corresponding to the lateral kinematic interaction factor. First, the comparison is made in terms of the foundation character
However, these differences between the soil-shell and soil-beam models ization variables: static stiffness, impedance functions and kinematic
10
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
Fig. 12. Validity of J for characterizing the relative foundation-soil stiffness in the computation of kinematic interaction factors.
interaction factors. A dimensionless analysis is made in order to present • Even above these limits, the foundation beam model accurately re
more general results. The main conclusions drawn from this study are: produces the OWT response in an acceptable frequency range for
seismic analyses. The upper bound corresponds to the frequency for
• The dimensionless parameter J proposed by Doherty et al. (2005) to which the foundation diameter coincides with half wavelength.
define the relative stiffness between the foundation and soil in static • This restrains the use of the beam model for extremely soft soils if the
can be also applied for dynamic analyses. high frequency content of the excitation is important. But for typical
• In general terms, the soil-beam model accurately reproduces the scenarios, the foundation beam simplification is a valid option for
global foundation response with respect to the rigorous soil-shell reproducing the OWT seismic response, making it a valuable tool
model. The agreement is quite good even for foundations with especially for design or optimization steps.
small aspect ratios.
• The accuracy of the beam model is reduced for high frequencies and CRediT authorship contribution statement
low foundation-soil stiffness contrast.
• Closed-form expressions are proposed in order to estimate the Guillermo M. Álamo: Conceptualization, Methodology, Software,
applicability range, in terms of maximum dimensionless frequency, Investigation, Writing - original draft. Jacob D.R. Bordón: Methodol
of the beam simplification. Those are functions of the foundation-soil ogy, Software, Investigation, Writing - original draft, Writing - review &
relative stiffness parameter J and depend on the foundation variable editing. Juan J. Aznárez: Supervision, Funding acquisition.
to compute and maximum admissible error.
Declaration of Competing Interest
An application example is also presented, in which the seismic
response of a 5 MW OWT including SSI effects is computed via a sub The authors declare that they have no known competing financial
structuring procedure. The FRF of key structural variables obtained by interests or personal relationships that could have appeared to influence
both models (shell and beam) are compared, and the main conclusions the work reported in this paper.
drawn from this example are the following:
Acknowledgements
• The soil-structure interaction effects significantly change the seismic
response of the OWT system. This work was supported by Agencia Estatal de Investigación (AEI) of
• The foundation variable whose modelling has more impact on the Spain through research project BIA2017-88770-R. J.D.R. Bordón is a
obtained results is the lateral impedance term, followed by the postdoctoral fellow of the ULPGC Postdoctoral Programme. The authors
lateral kinematic interaction factor. are grateful for this support.
• Below the proposed limiting frequencies, virtually the same results
are obtained regardless using the beam or shell model.
11
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107
The dimensionless parameter J was proposed by Doherty and Deeks (2003) to characterize the relative stiffness between the foundation and soil for
static analyses. In this appendix, its use in dynamic regime is tested by computing the impedance functions and kinematic interaction factors for
several configurations and comparing the results obtained for the same value of J.
Five shell thickness ratios t/D = {0.001, 0.01, 0.02, 0.05, 0.1} are combined with a continuum range of values for the ratio between the foundation
and soil shear modulus Gf /Gs in order to cover a comparable J interval. For brevity’s sake, only results for a configuration with aspect ratio L/D = 6 are
presented. The rest of dimensionless properties are equal to the ones defined in Section 3.
Fig. 11 plots the impedance functions against the parameter J for the studied configurations. Real and imaginary components are presented in pairs
for several dimensionless frequencies distributed in rows. Each column correspond to different impedance modes. The results show that the parameter
J can be used to represent the relative foundation-soil stiffness for the static and low frequency scenarios. For higher frequencies the results are sensible
to the thickness ratio if its value if larger than 2%, especially the vertical and lateral modes. However, the foundation elements for OWT structures
typically present thickness ratios below this value, so the parameter J can be safely used to represents the foundation-soil stiffness contrast.
Fig. 12 shows now the results in terms of kinematic interaction factors. The real and imaginary components of the lateral and rotational terms are
presented in pairs of rows. Each column correspond to different dimensionless frequencies (static values are omitted as their results are trivial). As
commented before, for thickness ratios below 5%, the behaviour of the different configuration is determined by the J parameter. Thus, it can be also
used to characterize the foundation-soil relative stiffness when studying the foundation kinematic response.
Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.compgeo.2021.104107.
12